Grossing up Family Expenditure Survey data for use in international accounts

In this paper we show how estimates of aggregate spending in the UK would be affected by using grossing weights that take account of the known dimensions of non-representativeness of the Family Expenditure Survey. Currently a uniform weighting scheme is used by the ONS. Having estiiuated a model of spending at different degrees of disameption, corresponding to the commodities of interest as far as ONS totals are concerned, we show which dimensions of non-response are most s4nificant in determining spending patterns. We derive five sets of grossing weights which control for combinations of these factors. We show the effects of these different grossing factors for commodity groups and for the goods for which the FES is the sole source of information in the National Accounts. The effects vag, but for some of these goods (for example, theatres and taxis) controlling for non-response in computing population aggregates could increase estimates of total spending by almost nine per cent.